Data and Analytics - Optum · Smart Measurement System electronic data capture and study management...

Data and Analytics: Preparing for the Future

Gregory Warren, FSA, MAAA, FCA Vice President, Actuarial Consulting Optum Consulting

2 Propriety and Confidential. Do not distribute.

Experience ‘in the room” where decisions are made

Context Strategy Tactics

We offer unique knowledge of the actuarial culture, roles, language, methods, and value drivers within various health care, risk-bearing entities.

Our offerings are ultimately designed to enhance the relevance and impact of strategies and tactics designed for risk-bearing stakeholders.

Proprietary and Confidential. Do not distribute.

3

Calculus-Based Statistical Theory Measuring Results Health Economic Impact Dealing with Uncertainty Model-Building Experts

Actuarial Science and Health Economics & Outcomes Research

Common Foundations

Law of Large Numbers (minimize statistical variation) Estimate Confounding Factors Financial Outcomes Book of Business Focus Identify Correlations Short/Intermediate-Term Horizons

Divergent Approaches and Applications

Building Bridges: Common Foundations, Divergent Approaches and Applications

Characteristic-Matched Studies (minimize confounding factors) Eliminate Confounding Factors Clinical & Economic Outcomes Disease Focus Identify Causations Intermediate/Long-Term Horizons

4

Optum actuarial solutions are integrated to an interdisciplinary approach that aligns strategy, value and execution across the product lifecycle

Discovery/preclinical Phase I Phase II Phase III Post launch/phase IV

MARKET ACCESS STRATEGY VALUE DEMONSTRATION STAKEHOLDER ENGAGEMENT

Que

stio

ns a

nsw

ered

Actuarial modeling and payer financial perspectives incorporated into traditional economic, clinical and pricing approaches

Define value landscape

• Current product performance?

• Payer, provider and patient unmet clinical needs?

• Payer addressable burden (risk stack) opportunity value?

• Potential sources of differentiation?

• Reimbursement requirements?

Measure impact and manage

• Actual outcomes and cost of care?

• Patient/provider uptake and gaps?

• Competitor product impacts?

• Value of services to stakeholders?

• How to assess/ manage financial risk?

• How do payers model value for future positioning?

Quantify economic value

• Data and study types?

• How are payers budgeting costs?

• How will payers model formulary decisions?

• Commercial value? • Clinical trade-offs? • Go/no-go

decision points? • Post-launch

evidence needs?

Develop value strategy

• Target patients? • Target populations? • Target sub-

populations? • Target value

statements? • Value thresholds? • Target claims

and endpoints? • Price/value/access

positioning?

Engage key stakeholders

• Stakeholder entities/individuals?

• Stakeholder issues/priorities?

• How to engage payer actuarial/ financial stake-holders?

• Tailor messaging? • Messaging

channels? • Resource needs? • When to use risk-

share agreements? • Clinical vs financial

risk agreements?

Support product launch

• Data for value proposition?

• How to quantify value in payer language?

• Reimbursement strategy?

• Pricing strategy? • Launch sequence? • Value of services

opportunities?

Pricing and reimbursement landscape

Ecosystem/market analysis

Filing and launch

Evidence strategy

Stakeholder engagement

Target value profile

Pricing analysis

Patient services

Evolving Healthcare Ecosystem Series (Actuarial Workshops)

Actuarial formulary design modeling

Health technology pipeline

Payer actuarial ROI modeling

Negotiation support

Risk-share agreement design

Risk-share pilot

Risk-share management

Actuarial formulary design modeling (Y2)

Actuarial payer addressable burden (episodes of care)

5

Del

iver

able

s

The Optum integrated strategic approach produces tangible actuarial deliverables for use in price, value, and access optimization


MARKET ACCESS STRATEGY VALUE DEMONSTRATION STAKEHOLDER ENGAGEMENT

Define value landscape

Competition • Competitive

analysis • Patient sub-

population analysis • Provider/patient

preference analysis • Market research • Evidence map

Reimbursement • Global payer review • HTA assessment • Treatment pathways • Site of care analysis • Benchmark

reimbursable file

Measure impact and manage

Value management • Market analytics • Value monitoring

and communication • Stakeholder value

map • Stakeholder value

realization plans • Market dynamics

and scenario planning

• Patient/provider services development, piloting and roll-out

• Risk-share management

Quantify economic value

Actuarial models • Payer pipeline cost

forecasts (HTP) • Payer formulary

design models • Payer addressable

burden analysis • Client ROI models

Evidence plans • Study designs • Integrated clinical/

evidence plans • HEOR and PRO

plans • post-launch plans

Develop value strategy

Target value profile • Value hypotheses • Stakeholder value • Target population • Active comparator • Target endpoints

Pricing analysis • Pricing value

concept • Early research • Horizon scanning • Market access

schemes • Value based

pricing • Initial projections


Engagement • Stakeholder

targeting • Hot-button analysis • Message

development • Regulatory review • Channel messaging • Training • Resource planning • Integrated value

Risk-share design • Design evaluation • Financial risk-

share design modeling

• Risk-share pilot


Launch plans • Final value

proposition • Reimbursement

strategy • Global and local

value dossiers • Global price corridor

and launch sequence

Post-launch plans • Label expansion • Target populations • New indications • Services strategy

6


EVIDENCE STRATEGY EVIDENCE DEMONSTRATION STAKEHOLDER ENGAGEMENT

Que

stio

ns a

nsw

ered

Actuarial modeling and payer financial perspectives incorporated into traditional economic, clinical and pricing approaches

Define evidence landscape

Measure real world impact

Quantify economic evidence

Develop evidence strategy



Natural history of disease

Evidence needs assessment

Disease registries

Prospective evidence generation

Evidence synthesis Updated

economic models

Budget impact modeling Safety and surveillance

Smart Measurement System electronic data capture and study management

Actuarial payer addressable burden (episodes of care)

Actuarial formulary design modeling

Health technology pipeline

Payer actuarial ROI modeling

Risk-share agreement

design

Early burden-of-illness

Systematic reviews

Registries

Post-launch prospective and retrospective studies

Risk-share management

Risk-share pilot

Actuarial formulary design modeling (Y2)

• Payer addressable burden (risk stack) opportunity value?

• What is the burden of illness?

• How does the disease progress and what is its history?

• Actual outcomes and cost of care?

• Evidence of financial impact of services to stakeholders?

• How to assess/ manage financial risk?

• How do payers incorporate evidence in their financial models?

• What is the value and affordability in a real-world setting?

• How are payers budgeting costs?

• How will payers model formulary decisions?

• Post-launch evidence needs?

• What is the value and affordability to payers, both local and globally?

• What are the clinical and economic trade-offs?

• How do payers think about the financial impact of the disease and how to manage it?

• Drivers of cost-effectiveness and required efficacy, safety, and price thresholds?

• What does the disease cost?

• Stakeholder issues/priorities?

• How to engage payer actuarial/ financial stake-holders with evidence?

• What evidence can best support decisions about risk-share agreements?

• How to quantify value in payer language?

• How to apply evidence to financial value propositions?

• What evidence to present and how best to present it?

• What is the impact on payer pharmacy budgets?

Evolving Healthcare Ecosystem Series (Actuarial Workshops)

Early phase economic models

Early phase and cost of illness model

Early phase and cost of illness model

Optum actuarial solutions are integrated to align with evidence and outcomes research execution across the product lifecycle

7

Del

iver

able

s


EVIDENCE STRATEGY EVIDENCE DEMONSTRATION STAKEHOLDER ENGAGEMENT

Define evidence landscape

Measure real world impact

Quantify economic evidence

Develop evidence strategy



Competition • Competitive analysis • Evidence map • Early phase

economic models • Burden-of-illness

models • Natural history of

disease studies

Evidence management • Updated evidence

supporting risk-share management

Economic models and reports based on real-world data • Submission to drug

plans and publications

Actuarial models • Payer pipeline cost

forecasts (HTP) • Payer formulary

design models • Client ROI models

Evidence plans

Economic Models • For AMCP dossiers,

Global HTA submissions

• Cost-effectiveness models

• Budget impact models

• Risk-benefit models

Early phase model and report • Cost-effectiveness

with threshold analyses

Cost-of-illness model and report • For publication and

value dossier development

Actuarial models • Payer addressable

burden models defining potential for total cost of care savings

Risk-share design • Design evaluation • Financial risk-share

design modeling • Risk-share pilot Updated Economic Models

Launch plans • Evidence

communication strategy

Post-Launch plans • Evidence update

strategy

The Optum integrated strategic approach produces tangible actuarial deliverables for use in evidence development and demonstration

Propriety and Confidential. Do not distribute.

Actuarial Review of Life Sciences Models

8

• Formal actuarial review of models to provide:

– Insight to how actuaries may use the models

– Detailed observations of model strengths and perceived gaps from an actuarial perspective

– Recommendations on the potential to modify the existing models to gain more actuarial acceptance

• May be applied to:

– Budget impact models

– Cost-effectiveness models

– Long-range planning financial models

Feedback How to Make Better

1) The BIM assume12 months per member - actuaries would not accept this assumption

Use actual member months instead of members

2) Actuaries would find it hard to accept “billed” or “acquisition” costs

Use allowed or paid costs

3) Health insurance companies consider costs and are structured based on books of business

Show the range results for Medicare, Medicaid, and Commercial populations

4) The ability to adjust the population are important for actuarial analyses

Add a table showing prevalence, utilization, and costs across enrollee demographics

5) Trend factors are very important for actuarial analyses

Explicitly include trend factors for prevalence, utilization, and costs


Actuarial Modeling

9

Our actuarial modeling engagements have begun as stand-alone “translational” projects but can now be seen as important sequential phases in assessing, building and delivering financial value propositions to actuarial and financial analytics decision-makers in health plans, PBMs and ACOs.

Payer ROI

modeling

Payer addressable

burden modeling

Actuarial go-to-market assessment

Formulary design

modeling

Risk-share agreements and other

contracting approaches


Payer Addressable Burden Analyses

Episodes of care define the opportunity for products to “bend the cost curve” based on the total costs of care and the pharmacy versus medical cost mix.

10

58.2%

41.8% $1,450$1,500$1,550$1,600$1,650$1,700$1,750$1,800$1,850

Costs of Primary Condition Total Patient Costs

Drivers of Cost Changes Severity and Costs


Formulary Design Modeling

Actuarial models similar to what is used by payers when determining formulary placement , medical drug benefits, and developing clinical program and utilization management policies . These modeling engagements provide insights to key levers that payers may use to exploit or mitigate financial risks.

11

Future State 1MAPD

Overall Market Overall Market

Year Company Direct Indirect Company Direct Indirect Total Scripts/1000 Revenue PMPM2017 4% 22% 74% $0.24 $3.12 $0.64 $4.00 119 $2.772018 5% 24% 71% $0.42 $4.50 $2.51 $7.43 144 $3.752019 5% 26% 69% $0.67 $6.36 $3.33 $10.36 169 $4.94

Plan Net Cost PMPMMarket Share

MarketScenario

Company

Overview Inputs Results

Results Saved Results

ReferencesReset All Defaults

Dictionary

Class Details Drug DetailsSummary Totals

Save Results

Summary

Model Assumptions 1. Eligible population will remain stable throughout years 2016 - 2018 2. Formulary status will remain consistent between 2017 and 2018 3. Four new drugs will enter in 2016 as specialty drugs 4. All drugs require prior authorization 5. New products will be excluded from formularies if the net costs are above market costs 6. Current drugs manufacturers will decrease rebates as exclusive contracts end after 2016 7. Pricing for new products is based on Product 1 AWP for all scenarios 8. The prevalence of Comorbidity X is 15% 9. Annual Utilization Trend – Expected to remain flat 10. Annual Unit Cost Trend – Remains flat due to current high costs of HCV therapies 11. Assumed average treatment of 12 weeks 12. Ultimate share is achieved in 2018


Client Return-On-Investment (ROI) Modeling

• Payer ROI modeling estimates the incremental financial “investment” and “return” of specific products relative to their competitors.

• Models for various client types utilizing “proxy client data” from Optum databases, with sensitivity testing of various population and assumption scenarios

12

InputsPerformance Measure Option 1Distribution LognormalGap Method 15+Confidence Interval 95%

Pharmacy Claim Cost 120% represented as Member Cost Share 100% represented as Formulary Rebate % 100% represented as

Number of Members 25000

`Calculations

High Average LowAverage competitor products' persistence (weeks): 14.53 14.53 14.53Average product persistence (weeks): 18.14 18.00 17.86Difference in persistence with product (weeks): 3.6 3.5 3.3

Annual Medical Claim Savings per Week of Persistence: $301 $303 $306Difference in Annual Medical Claims with Product: $1,087 $1,053 $1,018

Annual Therapy Class Average Pharmacy Claim Cost: $4,763 $4,763 $4,763Annual Therapy Class Average Member Cost Share: ($667) ($667) ($667)Annual Therapy Class Pharmacy Plan Cost: $4,097 $4,097 $4,097Annual Therapy Class Formulary Rebates: ($953) ($953) ($953)Annual Therapy Class Net Payer Cost: $3,144 $3,144 $3,144

Annual Product Pharmacy Claim Cost: $5,716 $5,716 $5,716Annual Product Member Cost Share: ($667) ($667) ($667)Annual Product Pharmacy Plan Cost: $5,050 $5,050 $5,050Annual Product Formulary Rebates: ($953) ($953) ($953)Annual Product Net Payer Cost: $4,097 $4,097 $4,097

Difference in Annual Net Payer Cost with Product: ($953) ($953) ($953)

Outputs

Annual Payer ROI with Product: 14% 11% 7%

Days Supply + 15Existing Payer Data

Contract guarantee

Payer “return”

Payer “investment”

Payer “ROI”

Risk Share Agreement AnalysisPayer ROI Model


Risk Share Strategies and Design

13

Structures, balancing risks/opportunities, selling, contracting, reconciliation and risk mitigation

1. Actuarially project future episode costs

2. Apply study outcomes to episode costs to calculate guaranteed savings

3. If actual savings are less, reimburse payer

4. If actual savings are more, share in the excess

Evolving Uses of “Real World” Data

David Dore, PharmD, PhD Vice President, Epidemiology – Optum Life Sciences Adjunct Assistant Professor – Brown University School of Public Health

Data and Analytics: Preparing for the Future


Medical Groups Hospitals

Integrated Delivery Networks

HOSPITALS

EMR1

EMR2

EMR3

Rx Platform

Billing System

EMR1

MULTI-SPECIALTY PRACTICES

Billing System

Rx Platform

Staging Area

Normalization & De-Identification Processes

• Medications • Lab Results • Vital Signs

• Physician Notes • Diagnoses • Procedures

• Demographics • Hospitalizations • Outpatient Visits

Humedica Business Model

SMALL GROUP PRACTICES

EMR2

Billing System

Rx Platform

PHYSICIAN OFFICES

EMR3

Billing System

Rx Platform

Humedica Comprehensive

Dataset


TARGETED CONTEXTUALIZATION OF SAFETY SIGNALS

What is the incidence of Adverse Event X in my RCT population?


Imagine this is your trial…

Propriety and Confidential. Do not distribute. 18


Mimicking the Trial Population

Trial Participantsn ~ 10,000

Observed StandardizedAge, Mean (SD) 64.3 (7.2) 64.9 (9.7) 64.9 (9.7)Gender (male), % 64.3 49.7 61.2Ethnicity, %

Hispanic or Latino 12.2 6.3 13.2Race, %

White 77.5 76.0 77.9Asian 9.9 1.8 9.0Black 8.3 11.1 8.3Other 4.3 11.1 4.8

Weight, kg, Mean (SD)* 81.8 (21.0) 73.9 (23.5) 82.1 (22.5)Body mass index, Mean (SD) 32.5 (6.3) 32.9 (7.4) 32.1 (7.0)Hyperlipidemia, % 77.0 54.0 77.6Coronary artery disease, % 56.8 19.2 57.4eGFR, %

<30 1.9 1.4 1.330-60 19.9 6.0 14.660-90 41.3 28.6 42.7>90 36.9 64.0 41.3

*Values edited to protect confidentiality

Baseline characteristics of participants in the Trial and patients in the Trial-like Cohort within the Humedica Research Database (01 January 2008 - 31 December 2013)

Trial-like Cohortn ~ 208,672


CHRONIC DISEASE EPIDEMIOLOGY

Does Hypoglycemia Increase Risk of Acute Cardiovascular Events?


First improve capture of hypoglycemia

0

2

4

6

8

10

12

14

2009 2010 2011 2012 2013

Prev

alen

ce (%

)

Year

Prevalence of hypoglycemia diagnoses or mentions, by year and identification method

ICD-9Algorithm

NLP Algorithm

CombinedAlgorithm*

Nunes AP, Yu S, Kurtyka KM, Senerchia C, Hill JM, Brodovicz KG, Radican L, Engel SS, Calvo SR, Dore DD. American Diabetes Association Annual Meeting 2014. San Francisco, CA. Abstract 5-LB. Available from: https://ada.apprisor.org/index.cfm?k=v7nb53u67w. Last accessed: 12 May 2015

https://ada.apprisor.org/index.cfm?k=v7nb53u67w


Interim Results, Nunes et al. EASD 2015 Oral Presentation

24

Any CVD Acute MI Full Cohort (n=82,321)

Events, N 39,209 1,649Unadjusted RR (95%CI)

Severe 2.02 (1.91-2.13) 1.92 (1.52-2.41) Mild-moderate 1.55 (1.08-2.21) 1.05 (0.17-6.31) Unknown 1.57 (1.54-1.61) 1.35 (1.19-1.52)

Adjusted* RR (95%CI) Severe 1.69 (1.59-1.79) 1.66 (1.31-6.30) Mild-moderate 1.43 (1.00-2.05) 0.96 (0.15-1.35) Unknown 1.30 (1.26-1.33) 1.19 (1.06-4.12)

Restricted Cohort, No baseline CVD (n=57,177)Events, N 17,179 626Adjusted** RR (95%CI)

Severe 1.77 (1.61-1.95) 1.18 (0.66-2.12) Mild-moderate 1.47 (0.90-2.39) 1.79 (0.48-6.72) Unknown 1.48 (1.42-1.54) 1.49 (1.24-1.81)

* Adjusted for age, gender, and prior CVD (any)** Adjusted for age and gender

HYP

OG

LYC

EMIA

SEV

ERIT

Y


OVERCOMING CLASSIFICATION CHALLENGES

Can we identify a cohort of patients with binge eating disorder?


Natural Language Processing for Binge Eating Disorder

26

• Binge eating: 2,919 • Binge eating disorder: 813 • Binge eat: 226 • Binge-eating: 104 • Binge eats: 80 • Binge eater: 62 • Bing eating: 9 • Binge-eat: 5

Observed Text Counts

• Binge eating: 3,389 • Binge eating disorder*: 829

Mapped Text Counts

Mapping of observed free text to unique medical concepts

Nunes et al. ICPE 2015 Abstract #941


Natural Language Processing for Binge Eating Disorder

27

Full BED Cohort (N=7,115)

Exact "Binge Eating Disorder"

Text Cohort (N=497)

N % N % Age at First Observed Diagnosis (years) 18-29 1,243 17.5 81 16.3 30-39 1,396 19.6 105 21.1 40-49 1,505 21.2 127 25.6 50-59 1,677 23.6 131 26.4 60+ 1,294 18.2 53 10.7 Sex Female 5,475 77.0 400 80.5 Male 1,640 23.0 97 19.5 Anthropomorphic Measures and Vital Status N Median N Median

Body Mass Index 5,739 35.6 368 38.5 Systolic Blood Pressure 6,760 122.0 454 122.0 Diastolic Blood Pressure 6,758 77.0 454 76.5 Heart Rate 1,163 78.0 66 77.0

Mortimer et al. ICPE 2015 Abstract #1028


REMEMBERING OUR ROOTS

De-identified Virtual Medical Records

Propriety and Confidential. Do not distribute. 29

Patient Information:Patient ID: 1, 2, … , k | Year of birth | Gender | Race | Index Date

Days from Index Date/ Timestamp

Visit / Encounter

ID (NLP Sequence)

Observation Type Description

=CONCATENATE("Date Specimen Collected:","

",DAT!I2,CHAR(10),"Date Result Available:", CHAR(10), DAT!F2,

CHAR(10)) =DAT!K2Laboratory

Result

=CONCATENATE("Test Name (local, mapped):", " ", DAT!AI2, ",", " ", DAT!AQ2, CHAR(10), "Specimen

Type (local):", " ", DAT!AC2, CHAR(10), "Result (local, normalized):", " ", DAT!AA2, ",", " ", DAT!AL2, CHAR(10), "Units (local, normalized):", " ", DAT!AE2,

",", " ", DAT!AJ2, CHAR(10), "Normal Range:", " ", "(",DAT!AM2, ")", CHAR(10), "Qualitative Test

Result:", " ", DAT!AK2)

Date Specimen Collected:

labres_collected_dateDate Result Available:labres_available_date

labres_enc_id

Laboratory Result

Test Name (local, mapped): name_m, standardcode_cui

Specimen Type (local): localspecimentypeResult (local, normalized): localresult_inferred,

normalizedvalueUnits (local, normalized): localunits, units_cui_m

Normal Range: (normalrange)Qualitative Test Result: mapped_qual_code


Scrambled Data Example

30

DateObservation

TypeDescription

Date Specimen Collected: 20AUG2013Date Result Available:

20AUG2013

Laboratory Result

Test Name: Calcium.totalSpecimen Type (local): Serum

Result (local, normalized): 9.40, mg/dLNormal Range: (8.9 - 10.3)

Qualitative Test Result: NormalDate Updated:

14JAN2015Date of Onset:

06SEP2014

Allergy Type of Allergy: DrugSpecific Allergen: Amoxicillin

Date of Diagnosis: 17AUG2011

Date Updated: 13JAN2015

Diagnosis Diagnosis Code: 272.4Diagnosis Description: Other or Unspecified Lipidemia

Present on Admission?: YESDiagnosis Active?: YES

Update Date: 11JAN2015

Observation Date: 29MAY2014

PAIN (Observation)

Value: 6Unit: out of 10

Update Date: 10JAN2015Result Date: 08MAY2014

Pulmonary Function Test

FEV1: 3.27FVC: 4.23

FEV1 - FVC Ratio: 77

Thank you David Dore 781.472.8458 [email protected]

Data and Analytics - Optum · Smart Measurement System electronic data capture and study management...

Documents

Transcript of Data and Analytics - Optum · Smart Measurement System electronic data capture and study management...